Artificial Intelligence (AI)-driven prototype for sex estimation from cranial Computed Tomography (CT) images: A multi-regional analysis of the orbits, mastoid process, and foramen magnum in a Thai population.

Sex estimation from skeletal remains is a fundamental step in forensic anthropology. However, traditional morphometric methods relying on linear measurements often suffer from inter-observer variability and limited diagnostic power. This study aimed to develop and validate an artificial intelligence (AI)-driven prototype using an object detection framework to estimate sex gender from Cranial Computed Tomography (CT) images in a Thai population. 3D reconstructed images of the skull focusing on three anatomical regions-Orbit, Mastoid Process, and Foramen Magnum-were utilized. Two AI models, one without augmentation (Roboflow 3.0 Base) and one with augmentation (Roboflow 3.0 Augmented), were developed using the Roboflow 3.0 Object Detection (Accurate) model with their performance compared against traditional morphometric analysis of craniometric indices. Performance was evaluated using an external validation set of 80 independent samples. The Roboflow 3.0 Base model showed superior performance with an overall accuracy level of 95.05%, significantly outperforming the traditional combined morphometric analysis, which yielded an accuracy level of just 34.11%. Furthermore, the object detection model generated precise bounding boxes, providing visual explainability for the predictions. The proposed AI prototype addresses several limitations associated with manual osteometry by effectively capturing complex non-metric traits. It serves as a practical assistive tool for forensic experts, offering objective evidence suitable for legal proceedings and holding potential for future scalability across diverse populations.